Programming with Python

He acts as a Senior Data Scientist and Data Architect specialized in Preprocessing and Feature Engineering with the Pandas library in Python. Your mission is to design a high-fidelity cleaning and transformation workflow for the **Nominal Categorical Variables Transformation** of the [[nombre_del_dataset]] dataset, focusing primarily on the [[columnas_a_transformar]] columns. These variables lack a logical order, so your proposal must be technical, avoiding the magnitude bias that a traditional Label Encoding would introduce, and prioritizing the semantic integrity of the data.


Start the process by performing an exploratory cardinality analysis for each column mentioned. If the cardinality is low, implement an optimized One-Hot Encoding approach using `pd.get_dummies` or preferably `OneHotEncoder` from Scikit-Learn to facilitate integration into pipelines. You must explicitly address the multicollinearity problem by removing one of the resulting columns (dummy variable trap) and configuring the system to handle unknown categories using the [[metodo_manejo_desconocidos]] parameter to avoid errors in the validation set.


For columns that have high cardinality (exceeding [[umbral_max_categorias]] single categories), do not use One-Hot Encoding to avoid curse of dimensionality. Instead, it proposes and implements a Binary Encoding or Frequency Encoding strategy. If you decide to opt for Target Encoding, you must include regularization or smoothing techniques to mitigate the risk of overfitting and data leakage, explaining how this method affects the relationship between the independent variable and the objective [[variable_target]].


The resulting code must be modular and structured under the best software engineering practices. Uses `ColumnTransformer` to encapsulate all transformations into a single reusable object. Be sure to include logic to group categories with a frequency of occurrence lower than [[porcentaje_minimo_frecuencia]]% under a generic label called 'Other', ensuring that the model is not distracted by statistical noise or outliers irrelevant to the overall prediction.


Finally, it generates a brief but rigorous technical documentation that compares the memory efficiency and performance impact of the models after applying these transformations. It includes a final validation block that prints the summary of the new columns created, the assigned data type and a verification that there are no null values ​​remaining after the transformation in the [[entorno_de_desarrollo]] environment.

If any key information needed to fill the bracketed fields is missing, ask me the necessary questions before answering.

He acts as a Software Architect and Relational Database Expert with over 10 years of experience in the Python ecosystem. Your goal is to design and implement a robust data architecture for a [Nombre_del_Proyecto] system that requires seamless technical management of complex relationships. I need you to develop the code necessary to handle a One-to-Many relationship between the entities [Entidad_Padre] and [Entidad_Hija], as well as a Many-to-Many relationship between [Entidad_Principal_M2M] and [Entidad_Asociada_M2M], using exclusively the [Libreria_Python_ORM] ORM.


For the One-to-Many relationship, ensure that each instance of the subordinate entity is obligatorily linked to its parent by precisely defining referential integrity constraints, cascading deletion behavior ([Politica_Cascade]), and index creation to optimize frequent search queries. The schema should allow a parent entity object to seamlessly access its collection of related objects using a backref property optimized to avoid the N+1 query problem using [Estrategia_Eager_Loading].


As for the Many-to-Many relationship, you must implement an association table (or bridge table) named [Nombre_Tabla_Asociacion]. This table should not be a simple union of IDs; It is essential that you include additional context fields such as [Atributo_Extra_Relacion] and [Marca_Tiempo_Creacion]. The code should explicitly reflect how to map this intermediate table to the selected ORM, allowing bidirectional navigation and insertion of data that includes these additional attributes in the relationship between [Entidad_C] and [Entidad_D].


The code provided should rigorously follow Python conventions (PEP 8) and use Type Hinting to improve readability and maintainability. Additionally, it includes a configuration section for connecting to the [Motor_SQL] engine and a detailed example of how to perform an initial migration. Also consider implementing validation methods within the models to ensure that business rules, such as [Regla_Negocio_Validacion], are met before any changes are persisted to the database.


Finally, generate a functional test script that executes a complete cycle: creation of seed records, establishment of relational links, persistence in the database, and a complex reporting query. This query must retrieve data from both relationships (1:N and M:M) simultaneously, applying a specific filter based on [Criterio_Filtrado_Reporte]. The end result should be a Python module ready to be integrated into a production-grade application, with explanatory comments in each logical block.

If any key information needed to fill the bracketed fields is missing, ask me the necessary questions before answering.